Diffusion Models, Synthetic Order Flow, and the Next Phase of Market Simulation

When Markets Become Generators

For decades, realistic market simulation has been one of the most elusive ambitions in quantitative finance. Traders wanted environments where strategies could be tested without hindsight bias. Regulators wanted sandboxes to explore counterfactual rules. Researchers wanted controlled worlds that still respected the brutal complexity of real markets. What they mostly got instead were replays of the past or simplified agent-based cartoons of reality. The TRADES framework changes that balance. By combining transformer architectures with conditional diffusion models, it pushes market simulation from imitation toward generation.